bvisible/mcp-ssh-manager

MCP SSH Server: 37 tools for remote SSH management | Claude Code & OpenAI Codex | DevOps automation, backups, database operations, health monitoring

406 ★47 forksJavaScriptUpdated 8d ago

What you need to know

MCP server that enables Claude Code and OpenAI Codex to manage multiple SSH connections — execute commands, transfer files, manage databases, create backups, monitor health, and automate DevOps tasks directly from the AI assistant. 37 tools, currently v3.7.0.

Install

npm install -g mcp-ssh-manager (or npx mcp-ssh-manager / clone + npm install)
claude mcp add ssh-manager node /path/to/mcp-ssh-manager/src/index.js (personal, project, or user scope)
Optional Bash CLI: cd cli && ./install.sh; configure first server with ssh-manager server add
Prerequisites: Node.js 18+, Bash 4.0+ (CLI), rsync (sync), sshpass (optional, rsync with password auth)

Usage

  • Configure servers in .env (SSH_SERVER_<NAME>_HOST/USER/KEYPATH/PASSPHRASE/PORT/PROXYJUMP/PLATFORM) or TOML for Codex (~/.codex/ssh-config.toml)
  • Ask the agent: 'List my SSH servers', 'Execute ls -la on production', 'Upload config.json to production:/etc/app/config.json'
  • Reduce context: ssh-manager tools configure (minimal mode = 5 tools, ~92% context reduction / ~40k tokens saved)
  • OpenAI Codex: ssh-manager codex setup / migrate / test; config in ~/.codex/config.toml

Key features

  • 37 tools in 6 groups: Core, Sessions, Monitoring, Backup, Database, Advanced
  • Auth: password, SSH key, ssh-agent (passphrase-protected keys), per-server platform=windows for Windows OpenSSH hosts
  • ProxyJump/bastion with chained multi-hop and circular-dependency detection; ProxyCommand for SOCKS5/custom proxies
  • Backup/restore (MySQL, PostgreSQL, MongoDB, files), health checks + alerts, DB management with SELECT-only query guarding
  • Per-server security modes (unrestricted / readonly / restricted) with an opt-in JSONL audit log; tool activation for context optimization
  • Config hot reload (no restart), persistent sessions, SSH agent forwarding (opt-in, defaults false)

Best for

Developers/SREs who want their coding agent to drive everyday server ops (execute, deploy, sync, backup, monitor) across many SSH hosts

Caveats

  • ssh_db_query is SELECT-only by design; ssh_db_* inputs are shell-quoted (command-injection fixes in v3.6.x)
  • SSH command timeout is advisory on Windows hosts — set PLATFORM=windows to skip the Linux timeout/sh -c wrapper
  • rsync password auth requires sshpass; key auth recommended
  • autoApprove is all-or-nothing per tool — use server security modes (readonly/restricted) when sharing the MCP with third-party agents or CI bots
  • SSH agent forwarding lets anyone with root on the remote host use your loaded keys for the connection lifetime — enable only for trusted servers
Platforms: Linux · macOS · WindowsClients: Claude Code · OpenAI Codex
Documentation ↗

Reviewed 2026-08-07

Topics

anthropicautomationbackupclaude-codedatabasedeploymentdevopserpnextfrappemcpmcp-servermcp-sshmodel-context-protocolmonitoringopenai-codexremote-serversshssh-managerssh-mcptoml
View on GitHub ↗
Stars
406★
Forks
47
Language
JavaScript
License
MIT
Created
2025-08-12
Last push
2026-07-29